摘要
为了满足相关匹配的要求,在原有的二元纯位相匹配滤波器的基础上,提出了改进的匹配滤波器算法,提出把二元纯位相匹配滤波器应用于人脸识别的理论和方法。简单介绍匹配滤波器算法和性能评价标准,详细阐述二元纯位相匹配滤波器设计方案,分析传统的二元纯位相匹配滤波器和优化的二元纯位相匹配滤波器的性能差别,对比二者在人脸识别中的性能特点。最后,给出不同角度二元纯位相匹配滤波器的仿真结果,给出匹配滤波器进行类内、类间识别的输出相关图,测试算法在FERET和AR shelter标准人脸库中的识别效果。结果证明这种新的人脸识别算法对有遮挡人脸识别效果较好。
In order to satisfy the requirements of the correlation matched,an improved approach based on Binary Phase-only Matched Filters（BPOMFs）was presented,and a face recognition algorithm based on BPOMFs was proposed.Firstly,the algorithm and performance evaluation criteria of matched filters were briefly introduced,and the filters design was detailed,and then the performance differences between traditional–BPOMFs and optimal–BPOMFs were analyzed,and the contrast between the two BPOMFs applied in face recognition was given.Finally,the simulation results of the two filters,and the correlation output of intra-person and extra-person are respectively given,and then the recognition rate on FERET and AR shelter face database are tested.The results show that this newly proposed face recognition algorithm outperforms the baseline algorithm in the obscure face recognition.
出处
《微计算机信息》
2010年第22期169-171,共3页
Control & Automation
基金
基金申请人:邹垚
项目名称:基于DM642智能人脸识别系统研制
基金颁发部门:上海市教委(JWCXSL0802)
关键词
人脸识别
相关匹配
二元纯位相滤波器
Face recognition
Correlation match
Binary phase-only filter
